--- Log opened Fri Jun 17 00:00:43 2016 | ||
-!- besser82 [~besser82@fedora/besser82] has joined #shogun | 01:02 | |
-!- mode/#shogun [+o besser82] by ChanServ | 01:02 | |
shogun-buildbot | build #21 of memleak - valgrind is complete: Failure [failed memory check] Build details are at http://buildbot.shogun-toolbox.org/builders/memleak%20-%20valgrind/builds/21 | 01:06 |
---|---|---|
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout] | 01:09 | |
lisitsyn | wiking: http://buildbot.shogun-toolbox.org/memcheck/20160616-2012.html | 01:13 |
lisitsyn | it seems I've bazdmegged out these leaks | 01:14 |
-!- GandalfTheWizard [~Eva@112.10.170.90] has joined #shogun | 03:02 | |
@wiking | \o/ | 03:02 |
-!- besser82 [~besser82@fedora/besser82] has quit [Ping timeout: 264 seconds] | 03:49 | |
-!- besser82 [~besser82@fedora/besser82] has joined #shogun | 04:55 | |
-!- mode/#shogun [+o besser82] by ChanServ | 04:55 | |
shogun-buildbot | build #22 of memleak - valgrind is complete: Failure [failed memory check] Build details are at http://buildbot.shogun-toolbox.org/builders/memleak%20-%20valgrind/builds/22 blamelist: Sanuj <sanuj.sharma.in@gmail.com>, Viktor Gal <vigsterkr@gmail.com>, Sergey Lisitsyn <lisitsyn.s.o@gmail.com>, Viktor Gal <viktor.gal@maeth.com>, Heiko Strathmann | 05:25 |
shogun-buildbot | <heiko.strathmann@gmail.com>, Saurabh7 <saurabh.mahindre@gmail.com> | 05:25 |
-!- besser82 [~besser82@fedora/besser82] has quit [Ping timeout: 246 seconds] | 05:25 | |
shogun-buildbot | build #1023 of nightly_none is complete: Failure [failed test] Build details are at http://buildbot.shogun-toolbox.org/builders/nightly_none/builds/1023 blamelist: Sanuj <sanuj.sharma.in@gmail.com>, Viktor Gal <vigsterkr@gmail.com>, Sergey Lisitsyn <lisitsyn.s.o@gmail.com>, Viktor Gal <viktor.gal@maeth.com>, Heiko Strathmann <heiko.strathmann@gmail.com>, Saurabh7 | 05:29 |
shogun-buildbot | <saurabh.mahindre@gmail.com> | 05:29 |
shogun-buildbot | build #1153 of nightly_default is complete: Failure [failed test notebooks] Build details are at http://buildbot.shogun-toolbox.org/builders/nightly_default/builds/1153 blamelist: Sanuj <sanuj.sharma.in@gmail.com>, Viktor Gal <vigsterkr@gmail.com>, Sergey Lisitsyn <lisitsyn.s.o@gmail.com>, Viktor Gal <viktor.gal@maeth.com>, Heiko Strathmann | 06:08 |
shogun-buildbot | <heiko.strathmann@gmail.com>, Saurabh7 <saurabh.mahindre@gmail.com> | 06:08 |
-!- sanuj [~sanuj@117.204.247.209] has joined #shogun | 08:05 | |
@wiking | arianepaola: https://sources.debian.net/src/shogun/3.2.0-7.3/ | 08:12 |
-!- sanuj [~sanuj@117.204.247.209] has quit [Ping timeout: 264 seconds] | 08:43 | |
-!- GandalfTheWizard [~Eva@112.10.170.90] has quit [Quit: Leaving.] | 09:06 | |
-!- sanuj [~sanuj@117.204.247.209] has joined #shogun | 09:10 | |
-!- sanuj [~sanuj@117.204.247.209] has quit [Read error: Connection reset by peer] | 09:23 | |
-!- sanuj [~sanuj@117.203.17.127] has joined #shogun | 09:38 | |
sanuj | lisitsyn, there? | 09:41 |
lisitsyn | sanuj: yes | 09:43 |
sanuj | lisitsyn, i'm leaving those which are causing error right now | 09:43 |
sanuj | most of the are templates | 09:43 |
sanuj | lisitsyn, like: template <class ST> class CFeatureSelection : public CPreprocessor | 09:43 |
lisitsyn | sanuj: yeah it is a bit difficult | 09:44 |
lisitsyn | sanuj: we'd need to discuss with HeikoS and wiking | 09:44 |
sanuj | lisitsyn, did you see how to do GaussianKernel from Kernel in swig? | 09:44 |
sanuj | and shall i start looking at plugins? | 09:45 |
lisitsyn | sanuj: not yet, we need a patch for class_list | 09:45 |
sanuj | okay | 09:46 |
-!- [Chris] [~Chris]@deadtime.informatik.uni-tuebingen.de] has joined #shogun | 09:46 | |
@wiking | lisitsyn: what? | 09:46 |
lisitsyn | wiking: what | 09:47 |
lisitsyn | wiking: bazdmeg! | 09:48 |
sanuj | :P | 09:49 |
@wiking | lisitsyn: what needs to be discused? | 09:55 |
lisitsyn | wiking: templates | 09:56 |
lisitsyn | wiking: we need to setup base classes | 09:56 |
@wiking | what's with them | 09:56 |
lisitsyn | and I am not sure what's with templates | 09:56 |
lisitsyn | say DenseFeatures<T> | 09:56 |
lisitsyn | do we export it for all T? | 09:56 |
@wiking | currently yes | 09:57 |
@wiking | afik | 09:57 |
@wiking | *afaik | 09:57 |
lisitsyn | wiking: have you seen my patch for meta cpp? | 09:57 |
lisitsyn | aka wrap(…)? :D | 09:57 |
@wiking | hehehe yeah | 09:58 |
@wiking | :) | 09:58 |
lisitsyn | wiking: the memleaks we have are openmp-related now | 10:00 |
-!- GandalfTheWizard [~Eva@112.10.170.90] has joined #shogun | 10:00 | |
-!- GandalfTheWizard [~Eva@112.10.170.90] has quit [Client Quit] | 10:04 | |
@wiking | lisitsyn: not all | 10:30 |
@wiking | there's some gaussian related stuff | 10:31 |
@wiking | that needs to be fixed still | 10:31 |
-!- sonney2k [~shogun@7nn.de] has quit [Ping timeout: 252 seconds] | 10:37 | |
-!- sonney2k [~shogun@7nn.de] has joined #shogun | 10:37 | |
-!- sanuj_ [~sanuj@117.203.17.127] has joined #shogun | 10:38 | |
-!- sanuj [~sanuj@117.203.17.127] has quit [Ping timeout: 240 seconds] | 10:44 | |
-!- GandalfTheWizard [~Eva@112.10.170.90] has joined #shogun | 10:55 | |
-!- GandalfTheWizard [~Eva@112.10.170.90] has quit [Quit: Leaving.] | 11:01 | |
-!- [Chris] [~Chris]@deadtime.informatik.uni-tuebingen.de] has quit [Read error: Connection reset by peer] | 11:12 | |
-!- sanuj_ [~sanuj@117.203.17.127] has quit [Ping timeout: 258 seconds] | 11:33 | |
-!- GandalfTheWizard [~Eva@112.10.170.90] has joined #shogun | 11:46 | |
-!- sanuj [~sanuj@117.203.17.171] has joined #shogun | 11:47 | |
-!- sanuj [~sanuj@117.203.17.171] has quit [Ping timeout: 276 seconds] | 11:57 | |
-!- c4goldsw [5da420e6@gateway/web/cgi-irc/kiwiirc.com/ip.93.164.32.230] has joined #shogun | 12:27 | |
c4goldsw | wiking Hey, you around? I'm still not able to build the most recent version of Shogun: https://gist.github.com/c4goldsw/2b7d1597d13472f7dd3bd92b014ed6ea | 12:28 |
-!- sanuj [~sanuj@117.203.17.171] has joined #shogun | 12:40 | |
c4goldsw | Hey sanuj, have a moment? | 12:46 |
sanuj | c4goldsw, yes, say | 12:46 |
c4goldsw | I'm still not able to build the most recent version of Shogun: https://gist.github.com/c4goldsw/2b7d1597d13472f7dd3bd92b014ed6ea . Have any ideas about how to go fixing this? | 12:46 |
sanuj | c4goldsw, did you run "make" before "make install"? | 12:48 |
c4goldsw | No - I just tried that and still received the same error | 12:48 |
sanuj | c4goldsw, what were your cmake flags? | 12:49 |
c4goldsw | I'm using ccmake, but I'll try making it with cmake and I'll report back. | 12:50 |
sanuj | c4goldsw, you need to do this | 12:50 |
sanuj | go to shogun dir | 12:50 |
sanuj | mkdir build | 12:50 |
sanuj | cd build | 12:50 |
c4goldsw | and then? | 12:50 |
sanuj | cmake -DENABLE_TESTING=on -DCMAKE_BUILD_TYPE=Debug .. | 12:50 |
sanuj | for dev build^ | 12:51 |
sanuj | make | 12:51 |
sanuj | and if you want to install it then => make install | 12:51 |
sanuj | or sudo make install | 12:51 |
sanuj | depending upon where you want to install it | 12:51 |
c4goldsw | sanuj: Do I call make from shogun or shogun/build | 12:54 |
c4goldsw | ? | 12:54 |
sanuj | shogun/build | 12:54 |
-!- besser82 [~besser82@fedora/besser82] has joined #shogun | 13:03 | |
-!- mode/#shogun [+o besser82] by ChanServ | 13:03 | |
c4goldsw | sanuj: when I run cmake -DENABLE_TESTING=on -DCMAKE_BUILD_TYPE=Debug .. , no build files are placed in shogun/build :/ | 13:04 |
c4goldsw | (and I run it from build) | 13:04 |
sanuj | c4goldsw, what is the output when you run the cmake command? | 13:04 |
c4goldsw | Hold on, placing into gist | 13:05 |
c4goldsw | https://gist.github.com/c4goldsw/ff9fca6febe9d080eb0eab896efba1a3 | 13:05 |
c4goldsw | sanuj^ | 13:09 |
sanuj | checking | 13:10 |
sanuj | c4goldsw, run "make" in shogun/build | 13:10 |
c4goldsw | I have - nothing returns. The directory is empty | 13:11 |
sanuj | okay | 13:11 |
sanuj | 2 mins | 13:12 |
sanuj | c4goldsw, it is because of this line | 13:13 |
sanuj | -- Build files have been written to: /home/chris/programming/shogun | 13:13 |
sanuj | try running "make" in shogun | 13:13 |
c4goldsw | That simply throws an error: | 13:14 |
c4goldsw | sanuj just refresh that gist | 13:14 |
c4goldsw | and go to the bottom | 13:15 |
sanuj | okay it won't work | 13:15 |
sanuj | your build files were written to shogun | 13:15 |
sanuj | they should be written to shogun/build instead | 13:16 |
sanuj | c4goldsw, did you make new "buid" dir in shogun and ran cmake with flags from shogun/build? | 13:17 |
c4goldsw | Alright, then cmake probably provides an argument for specifying where to output the build files. I'll look into to that and come back if I have any problems | 13:17 |
c4goldsw | Yes, I did | 13:17 |
sanuj | it should work without any argument | 13:17 |
c4goldsw | Unfortunately, it isn't :/ | 13:17 |
c4goldsw | sanuj: Thanks for your help though | 13:20 |
sanuj | c4goldsw, see https://gist.github.com/sanuj/93f2d43e2b1e40a10dc278748d3f5d0c | 13:23 |
sanuj | should be like ^ | 13:23 |
-!- GandalfTheWizard [~Eva@112.10.170.90] has quit [Quit: Leaving.] | 13:23 | |
sanuj | then you can do "make" and "make installl" from shogun/build as shown in the cmake output also | 13:23 |
sanuj | c4goldsw, if it persists, ask someone else | 13:24 |
-!- c4goldsw [5da420e6@gateway/web/cgi-irc/kiwiirc.com/ip.93.164.32.230] has quit [Quit: http://www.kiwiirc.com/ - A hand crafted IRC client] | 13:29 | |
-!- c4goldsw [5da420e6@gateway/web/cgi-irc/kiwiirc.com/ip.93.164.32.230] has joined #shogun | 13:55 | |
-!- sanuj [~sanuj@117.203.17.171] has quit [Ping timeout: 244 seconds] | 14:21 | |
-!- HeikoS [~heiko@nat-240-119.internal.eduroam.ucl.ac.uk] has joined #shogun | 14:50 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 14:50 | |
c4goldsw | Hey HeikoS, have a moment? | 15:05 |
@HeikoS | c4goldsw: yeah hi | 15:05 |
c4goldsw | I'm getting a build failure: https://gist.github.com/c4goldsw/2b7d1597d13472f7dd3bd92b014ed6ea . Any ideas on what I should do? | 15:05 |
@HeikoS | this is in a pr? | 15:06 |
@HeikoS | or where? | 15:06 |
c4goldsw | Ah, this is on my machine | 15:06 |
c4goldsw | I'm just running 'make' | 15:07 |
@HeikoS | c4goldsw: delete the parser output | 15:08 |
@HeikoS | your parse file is probably outdated | 15:08 |
@HeikoS | do git status and delete all files in the meta example generator dir | 15:08 |
@HeikoS | that should sovle it | 15:08 |
c4goldsw | Okay, I'll give that a try. Thanks. | 15:08 |
c4goldsw | HeikoS: Nothing from the generator dir shows up when I do git status, so I'll just stash what I've done and do a reset --hard against the main repo's develop branch | 15:12 |
-!- sanuj [~sanuj@117.203.17.171] has joined #shogun | 15:16 | |
@wiking | arianepaola: ping | 15:22 |
c4goldsw | sanuj: Hey, I managed to solve my problems: I had a cache file which was messing things up, so I just deleted it. Problem solved. | 15:24 |
c4goldsw | CMakeCache.txt specifically | 15:24 |
arianepaola | hello everyone | 15:26 |
arianepaola | hi wiking | 15:26 |
@wiking | arianepaola: https://sources.debian.net/src/shogun/3.2.0-7.3/ | 15:27 |
@wiking | or just apt-get source python-shogun | 15:27 |
@wiking | on debian will get you the files | 15:27 |
@wiking | for building the package | 15:27 |
arianepaola | thanks :-) | 15:28 |
arianepaola | can we add the sources to shogun on github? | 15:28 |
sanuj | c4goldsw, oh i see | 15:30 |
sanuj | didn't you delete the old build and use a fresh build dir? | 15:31 |
c4goldsw | Nothing was in build anyway, so I'm just building from there. | 15:31 |
c4goldsw | sanuj still getting different build errors, so I just fetched updates from the main repo and I'm rebuilding. | 15:32 |
sanuj | c4goldsw, what errors? | 15:33 |
c4goldsw | Generator file errors, but I'll let finish building again and see what happens. | 15:33 |
c4goldsw | from the meta example | 15:33 |
c4goldsw | examples* | 15:33 |
-!- [Chris]_ [~Chris]@deadtime.informatik.uni-tuebingen.de] has joined #shogun | 15:57 | |
[Chris]_ | hi, I have a question regarding combined kernels and cross validation. How should I specify the features for CrossValidation if I'm using a SVM with a combined kernel consisting of two rbf kernels with different feature encodings? | 15:57 |
c4goldsw | HeikoS: I'm still getting the same error, sadly. My develop branch is at the same level as shogun-toolbox/develop and I've rebuilt from scratch (after having deleted my build directory). Any ideas? | 16:14 |
c4goldsw | I'm building with 'cmake -DENABLE_TESTING=on -DCMAKE_BUILD_TYPE=Debug ..' | 16:15 |
sanuj | wiking, i need to discuss something about mkl | 16:16 |
sanuj | got time? | 16:16 |
c4goldsw | Rather, can I build Shogun without building the examples? I've ran into this issue before - I later fetched a more recent version of shogun from the main repository and it just started to work. | 16:19 |
sanuj | c4goldsw, i think HeikoS is not online right now | 16:19 |
c4goldsw | sanuj: ah, ok. | 16:20 |
sanuj | his away message: Ich bin gerade nicht hier | 16:21 |
c4goldsw | I'm just using an online IRC client, I can't see that info. | 16:21 |
c4goldsw | Should probably find a better one. | 16:21 |
sanuj | c4goldsw, i think you can see on an online IRC client also | 16:22 |
sanuj | right click / double click on his nick | 16:23 |
c4goldsw | sanuj: yep, you're right. | 16:23 |
c4goldsw | didn't know aobut that | 16:23 |
-!- HeikoS [~heiko@nat-240-119.internal.eduroam.ucl.ac.uk] has quit [Quit: Leaving.] | 16:35 | |
-!- HeikoS [~heiko@nat-240-119.internal.eduroam.ucl.ac.uk] has joined #shogun | 16:36 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 16:36 | |
@HeikoS | c4goldsw: worked? | 16:43 |
c4goldsw | HeikoS sorry was AFK. Nope | 16:44 |
c4goldsw | Can I build Shogun without building the examples? I've ran into this issue before - I later fetched a more recent version of shogun from the main repository and it just started to work | 16:45 |
-!- HeikoS [~heiko@nat-240-119.internal.eduroam.ucl.ac.uk] has quit [Quit: Leaving.] | 16:50 | |
Saurabh7 | whats the issue | 16:53 |
Saurabh7 | c4goldsw: just set them off in ccmake then | 16:53 |
c4goldsw | Saurabh7 Ah, ok | 16:54 |
-!- arianepaola_ [~ariane@unaffiliated/arianepaola] has joined #shogun | 16:57 | |
-!- sanuj [~sanuj@117.203.17.171] has quit [Ping timeout: 244 seconds] | 16:58 | |
-!- arianepaola [~ariane@unaffiliated/arianepaola] has quit [Ping timeout: 244 seconds] | 16:59 | |
c4goldsw | Saurabh7 Thanks! | 16:59 |
Saurabh7 | np | 17:00 |
-!- c4goldsw [5da420e6@gateway/web/cgi-irc/kiwiirc.com/ip.93.164.32.230] has quit [Quit: http://www.kiwiirc.com/ - A hand crafted IRC client] | 17:09 | |
-!- c4goldsw [5da420e6@gateway/web/cgi-irc/kiwiirc.com/ip.93.164.32.230] has joined #shogun | 17:10 | |
-!- HeikoS [~heiko@nat-240-119.internal.eduroam.ucl.ac.uk] has joined #shogun | 17:12 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 17:12 | |
-!- [Chris]_ [~Chris]@deadtime.informatik.uni-tuebingen.de] has quit [Read error: Connection reset by peer] | 17:12 | |
c4goldsw | HeikoS I've just disabled the building of meta examples, so it builds now. | 17:13 |
-!- OXPHOS [9d8b131c@gateway/web/freenode/ip.157.139.19.28] has joined #shogun | 17:16 | |
-!- sanuj [~sanuj@117.203.17.171] has joined #shogun | 17:23 | |
sanuj | HeikoS, yo | 17:24 |
sanuj | got time? | 17:24 |
@HeikoS | sanuj: jo | 17:24 |
@HeikoS | yeah | 17:24 |
sanuj | HeikoS, https://github.com/shogun-toolbox/shogun/pull/3294 | 17:25 |
@HeikoS | merged :) | 17:26 |
-!- shogun-notifier- [~irker@7nn.de] has joined #shogun | 17:26 | |
shogun-notifier- | shogun: Sanuj :develop * 24ab29c / doc/cookbook/source/examples/classifier/kernel_svm.rst: https://github.com/shogun-toolbox/shogun/commit/24ab29c1a2589c0f89af692d58b6ee88cf6737ed | 17:26 |
shogun-notifier- | shogun: minor correction in svm cookbook | 17:26 |
shogun-notifier- | shogun: Heiko Strathmann :develop * 8422681 / doc/cookbook/source/examples/classifier/kernel_svm.rst: https://github.com/shogun-toolbox/shogun/commit/8422681367b6dd25805dabd7a6d200b5d9996adb | 17:26 |
shogun-notifier- | shogun: Merge pull request #3294 from sanuj/cookbook_svm | 17:26 |
shogun-notifier- | shogun: | 17:26 |
shogun-notifier- | shogun: minor correction in svm cookbook | 17:26 |
shogun-buildbot | build #728 of trusty - libshogun - viennacl is complete: Failure [failed compile] Build details are at http://buildbot.shogun-toolbox.org/builders/trusty%20-%20libshogun%20-%20viennacl/builds/728 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com>, Sanuj <sanuj.sharma.in@gmail.com> | 17:27 |
@HeikoS | sanuj: what else? | 17:27 |
sanuj | HeikoS, i wanted to discuss about this | 17:28 |
sanuj | https://github.com/shogun-toolbox/shogun/pull/3250#discussion-diff-66655854R29 | 17:28 |
sanuj | if i try to init combined kernel without combined features | 17:28 |
sanuj | it gives error | 17:28 |
sanuj | SG_ERROR("CombinedKernel: Number of features/kernels does not match - bailing out\n") | 17:29 |
sanuj | HeikoS, the error is from here https://github.com/shogun-toolbox/shogun/blob/develop/src/shogun/kernel/CombinedKernel.cpp#L165 | 17:31 |
@HeikoS | checking | 17:31 |
@HeikoS | that seems like a good error messag eno? | 17:31 |
@HeikoS | do these match? | 17:32 |
@HeikoS | you can send a mini PR that improves the error message if you like | 17:32 |
@HeikoS | Where | 17:32 |
@HeikoS | Number of features (%d) and kernels (%d) does not match - bailing out\n") | 17:32 |
@HeikoS | Number of features (%d) and kernels (%d) does not match - bailing out\n") | 17:32 |
sanuj | HeikoS, okay | 17:32 |
@HeikoS | Number of features (%d) and kernels (%d) does not match\n | 17:32 |
sanuj | HeikoS, but i think this is a bug | 17:32 |
@HeikoS | remove bailing out and print number | 17:32 |
@HeikoS | where is the cpp listing that makes this happen? | 17:32 |
@HeikoS | can you send a gist? | 17:32 |
sanuj | HeikoS, let me explain the error | 17:33 |
sanuj | it arises because of precomputed custom kernel | 17:33 |
@HeikoS | I assumed that | 17:34 |
sanuj | HeikoS, and the error can be tracked if you see line 164 | 17:34 |
@HeikoS | yep | 17:34 |
sanuj | HeikoS, if i don't use combined features then for loop in line 79 is executed | 17:35 |
sanuj | so you get as many features as there are kernels | 17:35 |
sanuj | but f_idx++; happens for all kernels except custom kernel (see line 115) | 17:36 |
sanuj | HeikoS, do i make any sense? | 17:37 |
@HeikoS | sanuj: you just need to make sure it matches when custom kernels are used | 17:38 |
@HeikoS | it makes sense | 17:38 |
sanuj | okay | 17:38 |
@HeikoS | sanuj: can you fix it? | 17:39 |
@HeikoS | or need help with that? | 17:39 |
sanuj | HeikoS, so i added this condition: if (k->get_kernel_type() != K_CUSTOM) for the for loop body in line 79 | 17:39 |
sanuj | and it is working on my local | 17:39 |
sanuj | it's fixed | 17:39 |
sanuj | can i send a separate PR | 17:39 |
@HeikoS | yeah do that | 17:40 |
@HeikoS | nice! | 17:40 |
@HeikoS | question: | 17:40 |
@HeikoS | the previous example | 17:40 |
@HeikoS | had a custom kernel? | 17:40 |
@HeikoS | why did it work? | 17:40 |
@HeikoS | the one that you based yours on? | 17:40 |
sanuj | HeikoS, they used combined features | 17:41 |
sanuj | with one feature in it | 17:42 |
sanuj | and they had 2 kernels | 17:42 |
@HeikoS | why did it work? | 17:42 |
sanuj | out of which one was custom | 17:42 |
sanuj | this is the same thing as what i have done currently in the PR | 17:42 |
@HeikoS | because otherwise I would suggest you exactly replicate that? Or did I ask to change? | 17:42 |
sanuj | HeikoS, you asked to change | 17:42 |
sanuj | see if you look at my pR | 17:42 |
sanuj | then there are 3 kernels | 17:43 |
sanuj | one is custom | 17:43 |
sanuj | so there are 2 features in combined features | 17:43 |
sanuj | but if we don't want to use combined features | 17:43 |
sanuj | init() function implicitly creates combined features from line 79 for loop | 17:44 |
sanuj | and adds 3 features to the combine feature | 17:44 |
sanuj | so it screws up | 17:44 |
sanuj | HeikoS, got it? | 17:44 |
sanuj | i think it should be corrected, it's a bug | 17:44 |
@HeikoS | no I dont get it | 17:45 |
@HeikoS | the previous example | 17:45 |
@HeikoS | why did it work | 17:45 |
@HeikoS | ? | 17:45 |
sanuj | HeikoS, the previous examples had 2 kernels | 17:45 |
sanuj | out of which one was custom | 17:45 |
@HeikoS | and one feature object? | 17:46 |
sanuj | combined kernel was initialized by combined feature with one feature in it | 17:46 |
@HeikoS | you dont want to do that? | 17:46 |
@HeikoS | ah | 17:46 |
@HeikoS | and then implicitly creates a dummy feature inside? | 17:47 |
sanuj | no, custom kernel is precomputed so it does not need a dummy feature | 17:47 |
sanuj | but the custom kernel creates a dummy feature inside | 17:47 |
sanuj | when it is set with a kernel matrix | 17:48 |
@HeikoS | ok | 17:48 |
sanuj | HeikoS, let me summarize | 17:48 |
@HeikoS | I am sorry, sanuj, I dont understand what you want from me :D | 17:48 |
sanuj | last example: | 17:49 |
@HeikoS | nono | 17:49 |
@HeikoS | I get the xample | 17:49 |
@HeikoS | what is the problem? | 17:49 |
@HeikoS | I mean what is unclear? | 17:49 |
sanuj | HeikoS, do you understand the bug? | 17:49 |
@HeikoS | nope | 17:50 |
sanuj | okay so according to you | 17:50 |
sanuj | i have a combined kernel | 17:50 |
sanuj | with 3 kernels | 17:50 |
sanuj | one is custom | 17:51 |
sanuj | and i do | 17:51 |
sanuj | combined_kernel.init(features_train, features_test) | 17:51 |
sanuj | HeikoS, what should happen? | 17:51 |
@HeikoS | all features should be initialised with either the features provided, or dummy features for custom kernel | 17:51 |
sanuj | HeikoS, but if i have a precomputed custom kernel then? | 17:52 |
@HeikoS | nothing, should just check whether the dimensions match | 17:52 |
@HeikoS | I think dummy features is the way for that | 17:52 |
sanuj | HeikoS, when i do : combined_kernel.init(features_train, features_test) | 17:52 |
sanuj | it errors | 17:52 |
@HeikoS | ok fixit :) | 17:53 |
sanuj | HeikoS, i'll send a PR and then you can see what i mean :D | 17:54 |
@HeikoS | ok send it :) | 17:54 |
@HeikoS | let the code speak | 17:54 |
@HeikoS | OXPHOS: around? | 17:54 |
@HeikoS | arianepaola_: hi! | 17:55 |
@HeikoS | Saurabh7: hi! | 17:55 |
arianepaola_ | Hello HeikoS | 17:55 |
@HeikoS | arianepaola_: hi | 17:55 |
@HeikoS | arianepaola_: so, the cookbook counter is too low! | 17:55 |
arianepaola_ | looking into it HeikoS :-) | 17:55 |
@HeikoS | arianepaola_: good! | 17:56 |
@HeikoS | keep in mind for midterm, you want 6*2 at least | 17:56 |
@HeikoS | and as you see, it takes some time to iterate | 17:56 |
@HeikoS | as lots of issues on the way | 17:56 |
-!- arianepaola_ [~ariane@unaffiliated/arianepaola] has quit [Read error: Connection reset by peer] | 17:56 | |
@HeikoS | sanuj: any news on the swig stuff? | 17:57 |
sanuj | HeikoS, lisitsyn wanted to discuss something about that | 17:57 |
sanuj | with you and wiking | 17:57 |
@HeikoS | lisitsyn: what is it? | 17:58 |
@HeikoS | sanuj: you know what that was? | 17:58 |
sanuj | HeikoS, i'll update my swig branch and then we can talk in context to that | 17:59 |
@HeikoS | cool | 17:59 |
@HeikoS | sanuj: its time to get this finished | 17:59 |
@HeikoS | and start on plugins | 17:59 |
sanuj | i know | 17:59 |
OXPHOS | HeikoS: yes for now. I need to catch a flight soon sry. | 18:00 |
@HeikoS | OXPHOS: ok | 18:00 |
@HeikoS | OXPHOS: so need to discuss a few things | 18:00 |
@HeikoS | got a minute? Or should I write an email? | 18:00 |
OXPHOS | HeikoS: email is works better i guess | 18:01 |
@HeikoS | OXPHOS: ok, can I give you a primer here though? | 18:01 |
@HeikoS | OXPHOS: or do you need to leave like NOW ? :) | 18:01 |
-!- arianepaola [~ariane@unaffiliated/arianepaola] has joined #shogun | 18:02 | |
OXPHOS | HeikoS: go ahead! having brief lunch now | 18:02 |
-!- arianepaola [~ariane@unaffiliated/arianepaola] has quit [Excess Flood] | 18:02 | |
@HeikoS | OXPHOS: ok so | 18:02 |
@HeikoS | remember the difference between compile time and runtime we discussed last time? | 18:02 |
-!- arianepaola [~ariane@unaffiliated/arianepaola] has joined #shogun | 18:02 | |
arianepaola | yes HeikoS | 18:03 |
@HeikoS | arianepaola: ok cool! | 18:03 |
@HeikoS | Ill be away over the weekend mostly | 18:03 |
OXPHOS | HeikoS: yes | 18:03 |
@HeikoS | but I have some time now to merge things and give feedback | 18:03 |
@HeikoS | OXPHOS: ok so have a look here: https://github.com/shogun-toolbox/shogun/pull/3277/files/3baae8059acbc6fcf0423d2ef67dfd1b27c890b0#diff-4e89c8ea6b0b57d1f78952d268054c52R70 | 18:03 |
@HeikoS | OXPHOS: this hard-wires VIENNACL into the linalg vector at compile time. | 18:04 |
@HeikoS | which means GPU = VIENNACL | 18:04 |
@HeikoS | but we dont want that | 18:04 |
@HeikoS | we want to be able to choose at runtime between say VIENNACL and CUDA | 18:04 |
@HeikoS | so there cannot ever be any #ifdef checks in the linalf vector class | 18:04 |
@HeikoS | but just calls to virtual methods of some interface | 18:04 |
@HeikoS | of which the specialization then incliudes VIENNACL (and guards of course) | 18:04 |
@HeikoS | do you see the (important ) difference? | 18:05 |
OXPHOS | I think so, let me try to reiterate it - | 18:05 |
@HeikoS | pleas do :) | 18:05 |
OXPHOS | we can have several different GPUBackend.h /.cpp, which are wrapped by #ifdef HAVE_XX, and enables linalg calculation and memory transfer | 18:06 |
OXPHOS | or in one GPUBackend we have different ifdefs?? | 18:06 |
OXPHOS | so Vector just calls GPUBackend.transfer, with CPUptr | 18:07 |
OXPHOS | and GPUBackend decides which method to use | 18:07 |
@HeikoS | you you have a base class | 18:08 |
@HeikoS | GPUBackendBase say | 18:08 |
@HeikoS | which is purely abstract and has all the interfaces to to transfers | 18:08 |
@HeikoS | and then multiple specializations of it | 18:08 |
OXPHOS | okay so like the base vector/gpu vector before | 18:08 |
OXPHOS | i get it | 18:08 |
@HeikoS | with hard dependency on lib | 18:08 |
@HeikoS | and in your vector class, you just call the interface | 18:09 |
OXPHOS | sure | 18:09 |
@HeikoS | this way, no ifdef in your vector class | 18:09 |
@HeikoS | AND | 18:09 |
@HeikoS | more importantly | 18:09 |
@HeikoS | user can change between backends while shogun is running | 18:09 |
@HeikoS | that is the main point | 18:09 |
@HeikoS | of the whole refactoring | 18:09 |
@HeikoS | the rest is details | 18:09 |
@HeikoS | but this is the thing that this is about | 18:09 |
@HeikoS | try to really get your head around this :) | 18:09 |
-!- lambday [8028b10a@gateway/web/freenode/ip.128.40.177.10] has joined #shogun | 18:09 | |
-!- mode/#shogun [+o lambday] by ChanServ | 18:10 | |
OXPHOS | but the specialization/derived classes still need guards right? | 18:10 |
@HeikoS | OXPHOS: yeah sure | 18:10 |
@HeikoS | but these are just for the specialization | 18:10 |
@HeikoS | not for the interface itself | 18:10 |
@HeikoS | so if both CUDA and VIENNA are available at compile time, both are built | 18:10 |
OXPHOS | sure' | 18:10 |
@HeikoS | and then can easily be changed at runtime | 18:10 |
@HeikoS | OXPHOS: what lambday and I will do now, is to draft this, with some more attention on details | 18:11 |
@HeikoS | and then you can update your PR | 18:11 |
OXPHOS | and the gpuresgister - should it be an enum? | 18:11 |
@HeikoS | not sure I get what you mean | 18:11 |
OXPHOS | how do users specify which gpu backend to use if multiple is available? | 18:12 |
@HeikoS | there will be some method | 18:13 |
@HeikoS | that is a detail, we will put that in our draft | 18:13 |
@HeikoS | enum is ok | 18:13 |
OXPHOS | sure | 18:14 |
@HeikoS | have a good flight, will send something within the next hour | 18:14 |
OXPHOS | thx! | 18:14 |
OXPHOS | BTW could you check some PR if you still got time? | 18:14 |
OXPHOS | https://github.com/shogun-toolbox/shogun/pull/3287 | 18:14 |
OXPHOS | and the last comment of: https://github.com/shogun-toolbox/shogun/pull/3280 | 18:15 |
OXPHOS | HeikoS: thanks! ^ | 18:15 |
shogun-notifier- | shogun: lambday :develop * 5181c9b / / (3 files): https://github.com/shogun-toolbox/shogun/commit/5181c9b0c68c4ce56446b839381945f5004cfa7d | 18:16 |
shogun-notifier- | shogun: Added shift invariant kernel class. | 18:16 |
shogun-notifier- | shogun: Heiko Strathmann :develop * 784ae71 / / (3 files): https://github.com/shogun-toolbox/shogun/commit/784ae715d216a03f7439d99b6efc85e84b82634c | 18:16 |
shogun-notifier- | shogun: Merge pull request #3289 from lambday/develop | 18:17 |
shogun-notifier- | shogun: | 18:17 |
shogun-notifier- | shogun: added translational invariant kernel class, refactored distance kernel | 18:17 |
@HeikoS | OXPHOS: cheching | 18:17 |
shogun-buildbot | build #729 of trusty - libshogun - viennacl is complete: Failure [failed compile] Build details are at http://buildbot.shogun-toolbox.org/builders/trusty%20-%20libshogun%20-%20viennacl/builds/729 blamelist: lambday <heavensdevil6909@gmail.com> | 18:17 |
shogun-buildbot | build #730 of trusty - libshogun - viennacl is complete: Failure [failed compile] Build details are at http://buildbot.shogun-toolbox.org/builders/trusty%20-%20libshogun%20-%20viennacl/builds/730 blamelist: Heiko Strathmann <heiko.strathmann@gmail.com> | 18:18 |
-!- OXPHOS [9d8b131c@gateway/web/freenode/ip.157.139.19.28] has quit [Quit: Page closed] | 18:25 | |
sanuj | HeikoS, see here | 18:34 |
sanuj | https://github.com/sanuj/shogun/blob/tags-swig/src/interfaces/modular/SGBase.i#L556 | 18:34 |
-!- sanuj [~sanuj@117.203.17.171] has quit [Quit: Leaving] | 20:01 | |
-!- HeikoS [~heiko@nat-240-119.internal.eduroam.ucl.ac.uk] has quit [Quit: Leaving.] | 20:11 | |
-!- HeikoS [~heiko@nat-240-119.internal.eduroam.ucl.ac.uk] has joined #shogun | 20:11 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 20:11 | |
-!- HeikoS [~heiko@nat-240-119.internal.eduroam.ucl.ac.uk] has quit [Ping timeout: 258 seconds] | 20:15 | |
-!- c4goldsw [5da420e6@gateway/web/cgi-irc/kiwiirc.com/ip.93.164.32.230] has quit [Quit: http://www.kiwiirc.com/ - A hand crafted IRC client] | 20:22 | |
-!- c4goldsw [5da420e6@gateway/web/cgi-irc/kiwiirc.com/ip.93.164.32.230] has joined #shogun | 20:22 | |
-!- lambday [8028b10a@gateway/web/freenode/ip.128.40.177.10] has quit [Ping timeout: 250 seconds] | 20:30 | |
-!- sonne|osx [~sonne@x4db3bbe4.dyn.telefonica.de] has joined #shogun | 20:38 | |
-!- sonne|osx [~sonne@x4db3bbe4.dyn.telefonica.de] has quit [Quit: sonne|osx] | 20:53 | |
-!- shogun-notifier- [~irker@7nn.de] has quit [Quit: transmission timeout] | 21:17 | |
-!- sonne|osx [~sonne@x4db3bbe4.dyn.telefonica.de] has joined #shogun | 21:27 | |
-!- c4goldsw [5da420e6@gateway/web/cgi-irc/kiwiirc.com/ip.93.164.32.230] has quit [Quit: http://www.kiwiirc.com/ - A hand crafted IRC client] | 21:36 | |
-!- c4goldsw [5da420e6@gateway/web/cgi-irc/kiwiirc.com/ip.93.164.32.230] has joined #shogun | 21:37 | |
-!- sonne|osx [~sonne@x4db3bbe4.dyn.telefonica.de] has quit [Quit: sonne|osx] | 21:45 | |
-!- besser82 [~besser82@fedora/besser82] has quit [Ping timeout: 260 seconds] | 21:50 | |
-!- sonne|osx [~sonne@x4db3bbe4.dyn.telefonica.de] has joined #shogun | 22:10 | |
-!- c4goldsw [5da420e6@gateway/web/cgi-irc/kiwiirc.com/ip.93.164.32.230] has quit [Quit: http://www.kiwiirc.com/ - A hand crafted IRC client] | 22:24 | |
-!- sonne|osx [~sonne@x4db3bbe4.dyn.telefonica.de] has quit [Quit: sonne|osx] | 22:57 | |
--- Log closed Sat Jun 18 00:00:44 2016 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!